Research on health status assessment of power batteries for new energy vehicles combined with machine learning

The state of health (SOH) of the power battery of new energy vehicles is directly related to the safety and economy of the vehicle operation. Accurately evaluating SOH has important research and application value. This paper proposes a power battery SOH prediction model based on the CatBoost machine learning algorithm. Combining multi-dimensional operating characteristics such as capacity, voltage, current, and temperature, it realizes accurate identification of different battery degradation stages through feature engineering, window construction, and label regression. The experiment uses the lithium battery aging dataset publicly available in 2024 to evaluate the accuracy and stability of the model in multiple rounds of training and testing, and compares it with SVR and equivalent circuit models. The results show that the proposed model performs well in prediction error, model robustness, and the ability to identify different SOH intervals, verifying the adaptability of CatBoost under complex power battery conditions. The research results can provide reliable support for the intelligent and refined operation and maintenance of battery management systems.